Low- to moderate-dose thalidomide combination rescue after luspatercept failure in transfusion-dependent β-thalassemia: a real-world cohort study

Abstract

Background: A proportion of transfusion-dependent β-thalassemia (TDT) patients show suboptimal or heterogeneous response to luspatercept. We explored a structured rescue approach by adding low- to moderate-dose thalidomide after predefined luspatercept failure in real-world practice. Methods: This study was a single-center retrospective cohort study. Adults with TDT on luspatercept who met predefined criteria for suboptimal response (failure to achieve ≥ 33% reduction in RBC units over a 12-week window versus each patient’s baseline 12-week window, or failure to increase Hb by ≥ 1.0 g/dL under an unchanged transfusion trigger) received thalidomide 25–50 mg once daily while continuing luspatercept. Primary endpoint was 12-week transfusion independence (TI), defined as no transfusion for 12 consecutive weeks and Hb ≥ 9.0 g/dL. Secondary endpoints included ≥ 33%/≥50% reductions in RBC units over matched 12-week windows, longitudinal Hb change, and safety graded by CTCAE v5.0 and reported as exposure-adjusted incidence rates. Results: Eleven patients were included (median age 23 years; female 45.5%; β⁰/β⁰ 18.2%; prior splenectomy 63.6%). Twelve-week transfusion independence (TI) was achieved in 4/11 patients (36.4%; 95% CI 15.2–64.6). Proportions with ≥ 33% and ≥ 50% reductions in RBC units over matched 12-week windows were 45.5% (95% CI 21.3–72.0) and 45.5% (95% CI 21.3–72.0), respectively. Hemoglobin increased by 4.0 g/dL (Hodges–Lehmann, 95% CI 2.9–5.0; LMM p < 0.001). Total exposure was 112 patient-months; EAIR for any-grade treatment-emergent adverse events was 8.0 per 100 patient-months. No ≥Grade 2 neurotoxicity or thrombosis occurred. Conclusions: In TDT patients with predefined luspatercept failure, adding low- to moderate-dose thalidomide was associated with improvements in hemoglobin levels and reductions in transfusion burden with favorable tolerability. These findings warrant prospective validation of this structured rescue approach.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s00277-026-07100-2.

Introduction

Transfusion-dependent β-thalassemia (TDT) represents one of the most severe forms of thalassemia, characterized by ineffective erythropoiesis and lifelong dependence on regular red blood cell transfusions [1]. Allogeneic hematopoietic stem cell transplantation remains the only curative treatment; however, most adult patients are ineligible owing to age, comorbidities, or lack of suitable donors [1]. Consequently, novel therapeutic strategies are required to alleviate transfusion burden and improve long-term outcomes.

Recent advances in gene therapy and gene editing, including CRISPR-based approaches, have demonstrated the potential to achieve transfusion independence in selected patients [2, 3], although their accessibility remains limited in routine clinical practice.

luspatercept, a first-in-class activin receptor ligand trap, enhances late-stage erythroid maturation and has demonstrated clinical benefit in reducing transfusion requirements, as shown in the phase 3 BELIEVE trial [4]. However, responses remain heterogeneous across patients, particularly among patients with β⁰/β⁰ genotypes, leaving a proportion of patients transfusion dependent despite prolonged therapy [5–9]. In clinical trials, treatment discontinuation criteria were applied in patients with inadequate response or unacceptable toxicity, highlighting the need for effective management strategies in this setting [4].

Thalidomide, in contrast, has been shown to stimulate erythropoiesis and induce γ-globin expression, with randomized clinical trial evidence demonstrating improvements in hemoglobin levels and reductions in transfusion requirements in patients with transfusion-dependent β-thalassemia [10–14]. However, its use is limited by potential neurotoxicity, thromboembolic risk, and rare long-term adverse effects such as secondary malignancies, including leukemia [11, 13].

Despite these advances, there remains limited evidence to guide treatment escalation in patients with suboptimal response to luspatercept. Although combination approaches have been explored in small observational studies, including thalidomide-based combinations [15], standardized criteria for initiation, monitoring, and evaluation are still lacking [15, 16].

In this context, we explored a structured rescue approach in which low-dose thalidomide was introduced after predefined suboptimal response to luspatercept, with standardized criteria for treatment modification and outcome assessment. This strategy aimed to evaluate the feasibility and potential clinical benefit of combination therapy, including the achievement of protocol-defined transfusion independence and reduction in transfusion burden.

Rather than establishing a definitive treatment model, this study provides preliminary evidence to inform future prospective and multicenter investigations.

Methods

Study design

This study was a single-center retrospective cohort study.

Patients and luspatercept failure definition

Adult patients with transfusion-dependent β-thalassemia (TDT) receiving luspatercept were screened for eligibility. TDT was defined according to established criteria as requiring regular red blood cell transfusions (typically ≥ 8–10 units per year or ≥ 100 mL/kg/year) to maintain adequate hemoglobin levels, consistent with previous literature and TIF guidelines.

Eligible patients were those who met predefined criteria for suboptimal response and were treated at the First Affiliated Hospital of Guangxi Medical University between January 2024 and September 2025. All eligible patients during the study period were consecutively included.

Suboptimal response was defined a priori as either (i) failure to achieve a ≥ 33% reduction in packed red blood cell (RBC) units over a 12-week period compared with the individual baseline 12-week window, or (ii) failure to achieve an increase in hemoglobin (Hb) of ≥ 1.0 g/dL under stable clinical conditions, including an unchanged transfusion threshold (Hb 9.0 g/dL) and iron chelation regimen.

Exclusion criteria included patients who declined thalidomide treatment, had insufficient clinical data or follow-up, or had contraindications to thalidomide (e.g., uncontrolled thrombotic disease or severe peripheral neuropathy).

Rescue intervention

Patients continued luspatercept (starting dose 1.0 mg/kg with dose escalation according to the prescribing information) and received additional thalidomide at an initial dose of 25 mg once daily, with optional escalation to 50 mg based on clinical response and tolerability.

Antithrombotic prophylaxis was administered according to institutional practice (e.g., aspirin 100 mg once daily) unless contraindicated. Concomitant iron chelation therapy was allowed. Serum ferritin was measured using a chemiluminescent immunoassay at our center.

Endpoints

Primary endpoint: 12-week TI (no transfusion for 12 consecutive weeks and Hb ≥ 9.0 g/dL). Secondary endpoints: ≥33%/≥50% reductions in RBC units over matched 12-week windows; longitudinal Hb trajectories; time-to-first 12-week TI; safety per CTCAE v5.0 with exposure-adjusted incidence rates (per 100 patient-months).

Statistical analysis

Continuous data are summarized as median (IQR). Hb trajectories were analyzed with linear mixed-effects models (random intercept by patient; fixed effects for time, genotype, splenectomy, chelation, and dose tier). RBC units were modeled by negative-binomial regression with offset = log(time at risk). Binary outcomes (TI; ≥33%/≥50%) are presented with exact/Wilson 95% confidence intervals. Sensitivity analyses excluded follow-up < 12 weeks and early assessments ≤ 4 weeks. Two-sided α = 0.05. Analyses used R v4.5.1.

Results

Patients’ baseline characteristics

Among a total of 44 patients treated with luspatercept during the study period, 11 met the predefined criteria for suboptimal response and were included in the final analysis. The median age was 23 years; female 45.5%; β⁰/β⁰ 18.2%; prior splenectomy 63.6%; baseline Hb 6.8 g/dL; baseline ferritin 2895 µg/L; baseline HbF 42.8%. The median luspatercept dose was 1.13 mg/kg. Thalidomide started at 25 mg in 100%, and escalated to 50 mg in 63.6%; median thalidomide exposure 5 months.

Hemoglobin and iron parameters

After thalidomide was added to luspatercept, median Hb rose to 8.6 g/dL (range, 6.9–11.8). For patients followed > 3 months, Hb further improved to 9.6 g/dL (7.9–11.5). Median serum ferritin decreased from 2895 ng/mL (633–5469) at baseline to 2215 ng/mL (715–3976) after combination therapy.

Primary endpoint

Twelve-week TI occurred in 4/11 (36.4%; 95%CI 15.2–64.6). Median time-to-first 12-week TI was 10 weeks (IQR 8-12).

Transfusion burden (matched-window analysis)

Over baseline 12 weeks versus the first 12 weeks post-rescue, median packed-RBC units changed from 6 to 0 (negative-binomial rate ratio 0.18, 95%CI 0.09-0.37, p <0.001). ≥33% and ≥50% reductions were achieved in 5/11 (45.5%) (95% CI 21.3–72.0) and 5/11 (45.5%) (95% CI 21.3–72.0) , respectively.

Genotype-specific responses

Hb increases were observed across genotypes. Representative examples included patient 1 (β mutation with α-thalassemia trait), whose Hb rose from 6.9 to 10.9 g/dL, and patient 4 (β⁰/β⁰), from 8.3 to 11.8 g/dL. Notably, even patients with severe β⁰/β⁰ genotypes achieved clinically meaningful improvement, although case 9 (β⁰/β⁰) failed to sustain Hb elevation despite transient benefit.

Statistical analysis of hemoglobin changes

Paired t-tests demonstrated significant Hb improvement after thalidomide addition (mean difference + 1.98 g/dL vs. luspatercept alone, p = 0.001; +3.82 g/dL vs. baseline, p < 0.0001). Most patients achieved ≥ 1.0 g/dL Hb increase, ranging from − 1.2 to + 4.3 g/dL, with concomitant transfusion reduction of − 1 to − 14 units. A linear mixed-effects model confirmed a significant effect of treatment phase on Hb (p < 0.0001), with estimated means of 5.0 g/dL (95% CI 4.8–5.2) at baseline, 6.8 g/dL (95% CI 6.3–7.3) with luspatercept, and 8.8 g/dL (95% CI 8.1–9.4) after combination therapy. Neither time (p = 0.498) nor the treatment × time interaction (p = 0.190) were significant, indicating stable efficacy over 3–12 months. Median Hb increased by 4.0 g/dL at 12 weeks (Hodges–Lehmann, 95%CI 2.9-5.0; LMM p < 0.001). Spaghetti plots showed consistent upward trajectories from week 8 onward (Table 1).

Table 1.

Baseline characteristics

Characteristics	N = 11
Age, years	23 (18–49)
Sex, n (%)
Male	6 (54.5%)
Female	5 (45.5%)
Genotype, n (%)
β⁰/β⁰	2 (18.2%)
β⁺/β⁰	1 (9.1%)
β⁺/β⁺	2 (18.2%)
β⁰/Hb E	3 (27.3%)
β⁺/Hb E	3 (27.3%)
Baseline Hb, g/dL	6.8 (4.3–8.9)
Serum ferritin levels (ng/mL)	2895.00 (633.00–5469.95)
Splenectomy, n (%)
Yes	7 (63.6%)
No	4 (36.4%)
Myocardial iron deposition, n (%)
T2*>20 ms	10 (91%)
T2*<20ms	1 (9%)
Liver iron concentration (LIC), mg/g dry weight	19.7 (14.4–35.95)
Liver iron concentration (LIC), n (%)
Mild to moderate	4 (36.4%)
Severe	7 (63.6%)
luspatercept starting dose (mg kg⁻¹)	1.0
Thalidomide starting dose (mg), n (%)
25 mg	11 (100%)
Thalidomide Escalated dose (mg), n (%)
50 mg	7 (63.6%)
Antithrombotic prophylaxis, n (%)
Yes	4 (36.4%)
No	7 (63.6%)

Values are median (IQR) or n/N (%). Hb and transfusion trigger expressed in g/dL. LIC was expressed as mg/g dry weight. Severity categories were defined as: mild (< 7 mg/g), moderate (7–15 mg/g), and severe (> 15 mg/g)

Safety

Total exposure was 112 patient-months. EAIR for any-grade treatment-emergent adverse events was 8.0 per 100 patient-months ( 9 events in 5 patients). No ≥Grade 2 neurotoxicity or thrombosis occurred. Dose reductions in thalidomide occurred in 5 patients; no permanent discontinuation was required.

Discussion

In this single-center retrospective cohort, we explored a structured rescue approach for patients with transfusion-dependent β-thalassemia who did not achieve predefined response thresholds with luspatercept [4, 16]. Using a uniform primary end point—12-week transfusion independence (TI) defined as no transfusion for 12 consecutive weeks with Hb ≥ 9.0 g/dL—we observed TI in 4/11 patients (36.4%; 95% CI, 15.2–64.6) (Supplementary Fig. S2). The cumulative incidence estimates should be interpreted cautiously due to the small number of patients at risk over time. Consistent with this signal, hemoglobin increased after rescue; in our longitudinal model, the estimated marginal mean Hb rose from baseline to luspatercept monotherapy and further with combination therapy, aligning with the patient-level trajectories [16, 17] (Figs. 1, 2 and 3). In addition, matched 12-week windows confirmed clinically meaningful reductions in packed-RBC units after rescue (Table 2). Importantly, a 100% reduction in transfusion does not necessarily indicate sustained transfusion independence as defined by the study criteria. Taken together, these results support the feasibility of a structured, minimal-intervention rescue approach for patients who do not meet prespecified luspatercept response thresholds [18].

Fig. 1.

Individual hemoglobin trajectories across baseline, luspatercept monotherapy, and luspatercept + thalidomide rescue. Each thin line denotes one patient; bold markers indicate median (line) and IQR (band). The horizontal dashed line marks the unchanged transfusion trigger (Hb 9.0 g/dL). Wilcoxon signed-rank tests compare adjacent phases and report Hodges–Lehmann median differences with 95% CIs; linear mixed-effects models (LMM) for repeated measures are provided in Methods. Hb is expressed in g/dL (1 g/dL = 10 g/L)

Fig. 2.

Hemoglobin and transfusion responses after thalidomide addition compared with luspatercept alone. Most patients showed hemoglobin increases ranging from − 1.2 to + 4.3 g/dL, while transfusion requirements generally decreased by − 1 to − 14 units. A minority of patients exhibited minimal or no improvement

Fig. 3.

Swimmer plot of individual timelines spanning luspatercept monotherapy (pre-rescue) and luspatercept+thalidomide rescue. Each lane represents one patient. Pre-rescue (light blue) precedes rescue (light yellow), with week 0 marking rescue initiation. Symbols: ✚ indicates treatment-emergent adverse events (TEAEs); ↑/↓ indicates thalidomide dose escalation/de-escalation; ♦ indicates the first achievement of 12-week transfusion independence (TI), with the subsequent 12-week interval shaded in gray; ● indicates transfusion events. TEAEs of ≥Grade 2 neurotoxicity or thrombosis were not observed

Table 2.

Efficacy outcomes under the structured rescue

Outcome	n/N (%) or median (IQR)	Effect size (95%CI)	Notes
Primary TI (12-week TI, Hb ≥ 9.0 g/dL)	4/11 (36.4%)	—	Primary endpoint
≥ 33% reduction in RBC units (matched 12-wk)	5/11 (45.5%)	Risk difference + 45.5% (95% CI 21.3–72.0)	vs. baseline 12-wk
≥ 50% reduction in RBC units (matched 12-wk)	5/11 (45.5%)	Risk difference + 45.5% (95% CI 21.3–72.0)	vs. baseline 12-wk
RBC transfusion, baseline 12-wk → post-rescue 12-wk, units	6.0 (4.0–8.0) → 0.0 (0.0–2.0)	Rate ratio 0.18 (95% CI 0.09–0.37), p < 0.001	negative-binomial
Hb change from baseline (g/dL)	+ 4.0 (2.9–5.0)	HL Δ + 4.0 (95% CI 2.9–5.0), p < 0.001	LMM / Wilcoxon

“Matched 12-week” denotes each patient’s baseline 12-week window immediately preceding rescue versus the first 12-week window after rescue. Proportions are shown with 95% CIs (exact/Wilson). Rate ratios estimated via negative-binomial regression (offset = log time)

The limited single-agent effectiveness observed before rescue in our cohort is concordant with prior experience, including the BELIEVE trial, in which a substantial subset of TDT patients—particularly those with β⁰/β⁰ genotypes—fail to achieve clinically meaningful reductions in transfusion burden under luspatercept alone [5–9]. Our approach addresses this gap by predefining failure, standardizing when to intervene, and quantifying benefit within fixed windows, thereby reducing center-to-center variability in decision-making and facilitating external evaluation [15, 16].

Recent advances in gene therapy and gene editing have demonstrated the potential to achieve durable transfusion independence in selected patients with β-thalassemia [2, 3]. However, these approaches remain limited by accessibility, cost, and the need for specialized infrastructure, and are not yet widely available in routine clinical practice. In this context, optimizing pharmacologic strategies remains of practical importance.

Mechanistically, a complementary interaction is biologically plausible. Thalidomide has been reported to expand erythroid progenitors and induce γ-globin [10–13], supporting early erythropoiesis [19, 20], whereas luspatercept promotes terminal erythroid maturation by modulating TGF-β/SMAD2/3 signaling [6, 21]. Targeting distinct stages of erythropoiesis may mitigate ineffective erythropoiesis more effectively than either agent alone, which is consistent with the time course of Hb gains observed over the first 8–12 weeks following rescue [14, 17]. Although this study was not designed to resolve mechanistic causality, the temporal alignment between rescue initiation and standardized end-point responses is hypothesis-generating for prospective validation [16].

Similar combination or sequential treatment strategies have been explored in other hematologic disorders, where agents targeting different stages of erythropoiesis are used to achieve additive or complementary effects [19–21]. For example, in myelodysplastic syndromes and other anemia-related conditions, combination approaches have been employed to enhance erythroid response when single-agent therapy is insufficient [21].

Although direct evidence in β-thalassemia remains limited, these experiences from related hematologic settings provide indirect support for the rationale of combining agents with distinct mechanisms of action.

Safety was favorable. Only grade 1 treatment-related adverse events (TRAEs) were recorded, and no ≥grade 2 neurotoxicity or thrombosis occurred during the observed exposure. Expressing safety as exposure-adjusted incidence rates (8.0 events per 100 patient-months) enhances comparability across studies of differing durations and sample sizes [15] (Table 3), and is particularly relevant when using thalidomide, where dose-dependent toxicities are a concern (e.g., 100 mg/day). Our data support a low- to moderate-dose strategy (25–50 mg/day) with antithrombotic precautions and close monitoring as a practical balance between efficacy and tolerability [22, 23]. In addition, rare but potential long-term risks such as secondary malignancies, including leukemia, should also be considered.

Table 3.

Safety summary (CTCAE v5.0; exposure-adjusted)

Adverse event	Patients with ≥ 1 event, n/N (%)	Events, n	EAIR per 100 patient-months
Any TEAE	5/11 (45.5%)	9	8.0
Somnolence/fatigue (G1)	4/11 (36.4%)	4	3.6
Constipation (G1)	3/11 (27.3%)	3	3.2
Peripheral neuropathy (≥ G2)	0/11 (0%)	0	0
Thrombosis (≥ G2)	0/11 (0%)	0	0

Total exposure = 112 patient-months. EAIR = 100 × (number of events) /(patient-months). Antithrombotic prophylaxis: 4/11 (36.4%); EAIR = 3.6 events per 100 patient-months

Beyond the rescue phase, dose-adjustment strategies merit attention. For thalidomide, a low- to moderate-dose approach after hematologic improvement may help preserve response while limiting toxicity; abrupt discontinuation, by contrast, may lead to Hb decline and renewed transfusion requirements [17, 23]. For luspatercept, pragmatic stop-rules in responders remain undefined; current practice largely emphasizes dose holding for Hb > 11.5 g/dL rather than structured de-escalation [15, 18]. Dose adjustments, including escalation or reduction, were individualized based on tolerability and hemoglobin response. These observations underscore the need for prospective studies to evaluate maintenance and tapering strategies for each agent within a combined framework.

Collectively, we describe a structured rescue approach that may be applicable in clinical practice: (i) initiation after predefined luspatercept failure assessed over matched 12-week windows; (ii) introduction of low- to moderate-dose thalidomide with antithrombotic precautions and standardized monitoring; and (iii) evaluation using 12-week TI (Hb ≥ 9.0 g/dL) as the primary end point, accompanied by ≥ 33%/≥50% reductions in RBC units as key secondary measures. Standardizing these elements may improve consistency in clinical decision-making and facilitate comparison across studies [16, 18].

This study has limitations, including its retrospective design, small sample size, single-center setting, and limited follow-up, which constrain causal inference and generalizability. Nevertheless, the achievement of strict 12-week TI in a subset of patients who failed prespecified thresholds provides proof-of-concept that a structured, minimal-intervention rescue approach is feasible and may be clinically meaningful [17, 18]. Larger multicenter cohorts with pre-registered analytic plans, genotype stratification, and biomarker integration (e.g., HbF dynamics, SMAD-pathway readouts) are warranted to validate efficacy, optimize dose modulation, and establish de-escalation strategies for long-term management [16, 23].

Conclusion

In this study, we explored a structured rescue approach for transfusion-dependent β-thalassemia, in which low- to moderate-dose thalidomide was introduced after suboptimal response to luspatercept. A 12-week transfusion independence endpoint (Hb ≥ 9.0 g/dL) was used as an exploratory outcome.

Although the observed 36.4% transfusion independence rate in this small cohort should be interpreted with caution, this approach may represent a feasible option in real-world clinical practice. No ≥grade 2 neurotoxicity or thromboembolic events were observed. Further prospective, multicenter controlled studies are warranted to validate these findings.

Supplementary Information

Below is the link to the supplementary material.

Author contributions

Soulivanh Nanthachak, Cailan Wen and Rongrong Liu designed the study; Cailan Wen and Yumei Huang collected and analyzed the data; Soulivanh Nanthachak, Cailan Wen, and Yumei Huang drafted the manuscript; Zhaoping Gan, Lingyuan Pan cared for patients; Rongrong Liu and Yumei Huang conceived of the study, reviewed and revised the manuscript. All authors agreed in the final version.

Funding

The research was supported by Young Leader Talent Training Program of Guangxi Medical University (No. 202302), Advanced Innovation Teams and Xinghu Scholars Program of Guangxi Medical University, Joint Project on Regional High-Incidence Diseases Research of Guangxi Natural Science Foundation under Grant No. (2023GXNSFAA026293), NHC Key Laboratory of Thalassemia Medicine and Guangxi Key Laboratory of Thalassemia Research and the first-class discipline innovation-driven talent program of Guangxi Medical University.

Data availability

The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of the First Affiliated Hospital of Guangxi Medical University. Written informed consent was obtained from all participants.

Competing interests

The authors declare no competing interests.